JP4991596B2 - Optical waveguide circuit and multi-core central processing unit using the same - Google Patents

Description

The present invention relates to an optical waveguide circuit that propagates an optical signal using an optical waveguide, and more particularly to an optical waveguide structure at an intersecting portion in a circuit structure having portions where the optical waveguides substantially intersect each other.

  In recent years, research on silicon photonics, which can use mature process technology in silicon LSI, has enabled the realization of low-loss optical waveguides even in extremely fine and sharp bends. Miniaturization, low power consumption, and introduction and integration of optical wiring to silicon LSIs are becoming possible. As a candidate for such an optical waveguide, a silicon (Si) thin wire waveguide that can be formed on a silicon on insulator (SOI) substrate by a relatively simple technique is promising.

  It is important to improve the degree of integration of the optical waveguide circuit in order to keep costs down while responding to increasing signal volume and communication speed requirements. Therefore, it is necessary not only to arrange the waveguides in parallel, but also to provide locations that intersect each other. However, in an optical waveguide typified by a Si wire waveguide, signal light propagating in each waveguide is scattered, reflected, and interfered with each other at the intersection where two or more waveguides intersect each other. It is known that a large loss and crosstalk occur, and it has been difficult to design an optical waveguide circuit that requires a large number of intersections.

  In the conventional crossed optical waveguide, for example, two Si thin wire waveguides cross each other, and the signal light is propagated across the crossing portions. At this time, light scattering and reflection, mutual interference, and crosstalk occur at the intersection, and about 1.5 dB loss and -9.2 dB crosstalk occur as one signal light propagation characteristic (for example, Non-patent document 1). Since it has a point-symmetric structure, naturally, this large loss and crosstalk occur with respect to the propagated signal light in both intersecting waveguides.

  On the other hand, for example, a technique of reducing loss and crosstalk by a method of making the waveguide elliptical at the intersection is also known (see, for example, Non-Patent Document 1). In this structure, light scattering at the intersection is suppressed, and the loss of signal light propagating through the elliptical waveguide is about 0.1 dB, and the crosstalk is about −25 dB or less. It is possible.

  However, in this structure, the waveguide characteristics of the optical waveguide on the side where the elliptical structure is not formed are deteriorated compared to the conventional structure, resulting in high loss and high crosstalk. Moreover, when this elliptical structure is applied to both two waveguides intersecting and an attempt is made to reduce the loss of each signal light, the loss is still 1.2 dB, which is close to 1.5 dB when no countermeasure is taken. This is described in the non-patent document.

That is, with this structure, it is possible to improve the propagation characteristics of the optical waveguides on only one of the intersecting waveguides, but it is considered difficult to improve the propagation characteristics of both the two intersecting waveguides. In the crossed optical waveguide having such characteristics, it is difficult to realize high-density optical wiring integration that requires the formation of many crossing portions.
IEICE Transactions 2005/6 Vol. J88-C No. 6

It has been difficult to propagate not only signal light propagating through one of the waveguides but also all signal light propagating through the waveguide at a location where the optical waveguides intersect, with low loss and low crosstalk. If the signal light in all waveguides cannot propagate with low loss and low crosstalk, it is difficult to include intersections in the circuit design, leading to a lower degree of freedom in circuit design and integration of high-density optical wiring. It becomes an obstacle.

An optical waveguide circuit according to an embodiment of the present invention includes a lower clad layer formed on a substrate, a first optical waveguide formed on the lower clad layer, and the first optical waveguide delimited by the first optical waveguide. One end of the lower clad layer is formed with the front end facing the side surface of the first optical waveguide, and the front end is partitioned by the first optical waveguide and the second optical waveguide narrowed in a tapered shape A third optical waveguide having a tip formed opposite to the tip of the second optical waveguide on the other of the lower clad layers, the tip of the third optical waveguide being tapered. To do.

In the crossed optical waveguide structure according to the present invention, the loss and crosstalk of the signal light propagating in one waveguide can be reduced to almost zero, and the signal light propagating in the other intermittent waveguide also has almost zero crosstalk. Can be reduced. Therefore, it is possible to realize high-density optical wiring integration that requires the formation of many intersections.

    Details of the present invention will be described below with reference to the drawings.

    First, the guiding principle of the optical waveguide of the present invention will be described with reference to the drawings.

  In the present invention, as shown in FIG. 1, a lower clad layer 13 is formed on a substrate (not shown), and optical waveguides 7 and 8 are formed on part of the lower clad layer 13. Although the optical waveguides 7 and 8 form intersections that intersect each other, the optical waveguides 7 are not in direct contact because they are intermittent at the intersections. The tips of the intermittent portions of the optical waveguide 7 have taper structures 9 and 10 that taper toward the intersections. An upper cladding layer 14 having a refractive index higher than that of the lower cladding layer 13 is formed on the taper structures 9 and 10. According to the present invention, since the signal light 11 propagating through the optical waveguide 7 is gradually weakened in the tapered structure portion 9 by adopting such a crossed waveguide structure, the refractive index is reduced. Cannot be propagated through the large optical waveguide 7 and is partially coupled to the lower cladding layer 13 having a low refractive index. Thereafter, the signal light 11 propagates through the lower clad layer 13, but the upper clad layer 14 having a higher refractive index than the lower clad layer 13 is formed in the upper portion along the direction in which the signal light 11 is desired to propagate. 13 is guided to propagate in parallel to the optical waveguide 7 without diverging. The signal light 11 propagating in the lower clad layer 13 passes through the waveguide 7 which is a core layer having a high refractive index in the tapered structure portion 10 where the optical confinement effect gradually increases again after passing over the intersection. Partially combined. In this way, the signal light 11 can propagate through the optical waveguide 7 again beyond the waveguide intersection without directly contacting the optical waveguide 8.

  As described above, according to the present invention, the signal light is once moved to another layer, so that the state where two kinds of signal light directly overlap and influence each other at the intersection is successfully avoided. For this reason, the waveguide 8 can have a structure that is not different from the normal linear waveguide structure even at the intersection, and as a result, the signal light 12 propagating in the waveguide 8 is lost at the intersection. And almost no crosstalk.

  On the other hand, with respect to the signal light 11 propagating through the intermittent optical waveguide 7, loss due to scattering and reflection of the signal light at the intersection can be greatly reduced, and crosstalk due to interference and crosstalk between the signal lights is also dramatic. Can be reduced. However, at this time, the loss due to the interlayer movement of the signal light 11 at this intersection occurs as a new loss that was not found in the conventional method. The loss due to the interlayer movement to cross the intersection is about 1.5 to 2 dB, which is almost the same as the value of 1.5 dB in the prior art. Therefore, when the propagation characteristics of the crossing portion of the propagation signal light 12 in the optical waveguide 7 are compared with the characteristics of the prior art, the loss is substantially equal and the crosstalk can be reduced to almost zero.

    As described above, in the crossed optical waveguide structure according to the present invention, loss of one signal light and crosstalk can be reduced to almost zero, and the other signal light can also propagate with almost zero crosstalk. It can be said that it is a very excellent structure compared with.

  The details of the present invention will be described below with reference to the drawings by giving specific embodiments.

(First embodiment)
FIG. 2 is a perspective view for explaining the outline of the crossed optical waveguide structure according to the first embodiment of the present invention, and FIG. 3 is a sectional view thereof.

A lower clad layer 16 made of a SiO 2 film is formed on a Si substrate 15, and optical waveguides 17 and 18 made of Si are formed on a part of the lower clad layer 16. Signal light 22, signal light is contained in each of them. 23 is propagated. The thickness of the lower cladding layer 16 is 3 μm, and the optical waveguides 17 and 18 are 250 nm thick and 450 nm wide. The optical waveguides 17 and 18 form intersections that intersect each other, but are not in direct contact because the optical waveguide 17 is intermittent at the intersections. The tips of the intermittent portions of the optical waveguide 17 have taper structures 19 and 20 that taper toward the intersection, respectively. The length of the taper structures 19 and 20 is 200 μm, and the width of the taper tip portion is 80 nm. The tip portions of the taper structures 19 and 20 are arranged at a distance of 50 μm from the side wall of the optical waveguide 18. At this time, in FIG. 2, the angle at which the optical waveguides 17 and 18 intersect is illustrated as 90 degrees for the sake of clarity, but it is not necessarily required to be 90 degrees. The reflection or crosstalk of the signal light at the crossing portion as seen in the conventional structure does not occur in the optical waveguide crossing structure of the present invention. For this reason, loss and crosstalk increase / decrease are not caused by the difference in the crossing angle, and the crossing angle can be freely designed. However, if the crossing angle is too steep and the optical waveguides are joined at a portion other than the crossing portion, unnecessary loss and crosstalk are caused. From the above, the crossing angle of the optical waveguides 17 and 18 is preferably 5 to 90 degrees. An upper cladding layer 21 having a refractive index higher than that of the lower cladding layer 16 is formed on the taper structures 19 and 20. It is desirable that the refractive index and thickness of the upper cladding layer have the following relationship. That is, when the refractive index of the upper cladding layer is n ₁ , the thickness is d ₁ , the thickness of the optical waveguide is d ₂ , and the wavelength of incident light is λ,
d ₂ ≦ d1 ≦ λ / n ₁ (1)
It is desirable that At this time, the light does not propagate in the upper cladding layer but is coupled to the optical waveguide. Here, polyimide having a refractive index of 1.53 is used as a material. The upper cladding layer 21 has a thickness of 600 nm, a width of 3 μm, and a length of 300 μm in order to include the tapered structure portion. Note that the tip portion of the upper clad layer 21 on the intersecting portion side can have a tapered structure (not shown) in the same manner as the optical waveguide. Note that FIG. 2 shows only the intersections of the optical waveguides, and light emitting, light receiving elements, other various devices, wirings, and the like are formed in other regions to which the waveguides are connected.

Such a structure can be realized by using a silicon on insulator (SOI) substrate. That is, the lower cladding layer 16 is a buried insulating film (SiO ₂ film) of the SOI substrate, and the optical waveguides 17 and 18 are obtained by processing the Si layer of the SOI substrate into a thin line shape.

  As described above, by adopting such a crossed optical waveguide structure, the optical waveguide 18 can take almost the same structure as the straight waveguide structure even at the intersection, and propagates through the optical waveguide 18. In the signal light 23, almost no loss or crosstalk occurs at the intersection, and the signal light 23 can be reduced to almost zero.

On the other hand, the signal light 22 propagating in the optical waveguide 17 is temporarily distributed and coupled into the lower cladding layer 16 in order to weaken light confinement in the optical waveguide 17 having a large refractive index through the taper structure. By passing through the lower cladding layer 16, the intersection is crossed. The thickness d ₀ of the lower cladding layer is such that the wavelength of light guided into the optical waveguide 17 is λ, and the refractive index of the lower cladding layer is n ₀ .
d ₀ ≧ λ / n ₀ (2)
It is desirable that

  As described above, since the upper clad layer 21 having a refractive index larger than that of the lower clad layer is formed on the upper portion of the optical waveguide 17 in parallel with the optical waveguide 17, the signal light 22 propagating through the lower clad layer is formed. The traveling direction is guided in a direction parallel to the optical waveguide 17. However, the refractive index of the upper cladding layer 21 needs to be increased so that the light incident on the lower cladding layer 16 is sufficiently strongly attracted to the upper cladding layer 21. In this embodiment, the tip of the tapered structure portion of the intermittent waveguide 17 is arranged at a distance of 50 μm from the other waveguide 18 that intersects. This distance may be 30 to 200 μm, but if the distance is shorter than 30 μm, the signal light 22 reaches the other waveguide 18 that intersects before the signal light 22 finishes being distributed and coupled to the lower cladding layer. Propagation loss will increase. On the other hand, if the distance is larger than 200 μm, the distance that the signal light 22 is propagated in the lower clad layer 16 becomes longer, the components scattered in the lower clad layer 16 increase, and the loss increases. When the distance is 50 μm, the loss is minimized, which is preferable.

  Further, when the tip portion of the upper clad layer 21 on the intersecting portion side has a taper structure, it is possible to suppress the diffusion of the signal light 22 into the lower clad layer and the propagation loss due to scattering at the intersecting portion. At this time, the length of the tapered portion of the upper cladding layer is preferably in the range of 100 μm or less. Further, it is more preferably about 20 μm. This is because if the length of the tapered portion is too long, the optical confinement effect on the signal light propagating through the waveguide becomes too weak, and the propagation efficiency decreases. The signal light 22 beyond the intersecting portion is coupled again to the optical waveguide 17 having a large refractive index through the taper structure. At this time, in FIG. 2, the tip portions of the tapered waveguide taper structures 19 and 20 are located on the same axis as the z axis when the traveling direction of the signal light is the z direction. However, since the signal light actually propagates in the lower cladding layer, it is not necessary that both end portions are strictly positioned in the front, and if it is about 300 μm, loss increases even if it is displaced from the front. Never do. However, if the deviation is more than that, the loss increases, which is not preferable. Similarly, the refractive index of the upper clad layer 21 is increased so that the signal light 5 propagating through the lower clad layer 16 is sufficiently attracted to the upper clad layer 21 and the tapered portion of the optical waveguide 17. In this case, although light is not confined in the upper clad layer 21, the signal light 22 propagating through the lower clad layer 16 is attracted to the upper clad layer 21 side, and the mode matching of the coupled light in the tapered structure portion is performed. Play a role to help.

  FIG. 4 shows a state of light propagation of the signal light 22 in the crossed optical waveguide structure. FIG. 4A is a diagram showing the light distribution with respect to the thickness direction (Y direction) when the optical coupling device according to this embodiment is viewed from the side. And the optical coupling efficiency in the light traveling direction (Z direction). FIG. 4B is a diagram showing a light distribution with respect to the width direction (X direction) when the optical coupling device is viewed from above. FIG. 4C shows the optical coupling efficiency in the light traveling direction (Z direction). The white part in the figure shows higher light intensity. From these figures, the signal light 22 propagating through the optical waveguide 17 is temporarily transferred to the lower cladding layer 16 immediately before the intersection, and is guided in the direction parallel to the optical waveguide 17 by the upper cladding layer 21 to be in the lower cladding layer 16. It can be clearly seen that after passing through and crossing the intersection, it returns to the waveguide 17 again. Further, the loss at the intersection is about 2 dB, which is about the same as the value of the prior art. It is considered that most of the loss here is due to a component that diffuses into the lower cladding layer 16 and a component that scatters at the intersection.

  At this time, the lower cladding layer 16 is dug down in parallel with both sides of the intermittent optical waveguide 17 to form a trench region (not shown), so that the signal light 22 in the lower cladding layer 16 is formed. It is also possible to suppress diffusion and reduce loss. In this case, the trench region may be located outside the width of the upper cladding layer 21. In the present embodiment, the width of the upper clad layer 16 is 3 μm, and therefore, it is only required to be 1.5 μm or more away from the tapered portion of the intermittent waveguide 17. However, since the effect of suppressing diffusion becomes smaller as the distance increases, the distance is preferably 1.5 to 2.5 μm, and more preferably 1.5 μm. In this case, the propagation loss at the intersection of the signal light 22 can be reduced to about 1.5 dB.

  At this time, the crosstalk at the intersection of the signal light 22 is -60 dB or less, which is almost zero.

  Thus, in the crossed optical waveguide structure according to the present invention, the loss and crosstalk of the signal light propagating in one waveguide can be reduced to almost zero, and the signal light propagating in the other intermittent waveguide is also cross-linked. Talk can be reduced to almost zero.

(Second Embodiment)
1 shows an embodiment of an optical waveguide circuit to which an optical waveguide intersection according to the present invention is applied. FIG. 5 is a schematic diagram of a multi-core central processing unit in which six arithmetic cores are connected by an optical waveguide circuit. In the figure, parts not related to the present invention, such as signal transmission / reception ports and electrodes to the outside, are omitted. Such multi-core arithmetic elements have already been put into practical use as long as the communication between the cores is electrical wiring.

  The multi-core chip is formed using an SOI substrate, uses a SiO2 layer as a lower cladding layer, and forms an Si waveguide by processing the upper Si layer to form an optical waveguide. The optical signal transmission / reception ports 24 on the arithmetic cores 1 to 6 are connected to optical signal transmission / reception ports on all other arithmetic cores via the optical waveguide 25, and communication between all the multicores is possible. At this time, in this embodiment, the optical waveguide crossing portions 27 are generated at approximately nine places. In particular, when attention is focused on the optical waveguides from the arithmetic cores 1 to 6, 3 to 4, there are four optical waveguide intersections. In these optical waveguide intersections, in the case of the conventional technique, as described above, a loss of 1.5 dB and a crosstalk of −9.2 dB occur in each of the two signal lights that intersect at each intersection. . For this reason, the signal light from the cores 1 to 6 and 3 to 4 passing through four intersecting portions causes a great loss and crosstalk, and accurate communication becomes difficult. For this reason, a multi-core chip using an optical wiring by applying a conventional structure has not been put into practical use yet. On the other hand, as shown in the figure, when the crossing structure according to the present invention is applied to the crossing portion of the optical waveguide, the crosstalk can be reduced to -60 dB or less, and the overall loss is suppressed to half or less of the conventional structure. Is possible.

In addition, this invention is not limited to embodiment mentioned above. Although the SOI substrate is used in the embodiment, the SOI substrate is not necessarily used. Any structure may be used as long as the optical waveguide is disposed on a part of the lower cladding layer and the upper cladding layer is provided thereon. Furthermore, the materials of the upper clad layer, the lower clad layer, and the optical waveguide are not limited to the embodiment, and can be appropriately changed according to the specifications. When the refractive index n _{1 of} the upper cladding layer, the refractive index n ₂ of the lower cladding layer, and the refractive index n ₀ of the optical waveguide,
n ₂ <n ₁ <n ₀ (3)
If it is the relationship.

In addition, the length, thickness and width of the upper cladding layer, the thickness of the lower cladding layer, the thickness and width of the optical waveguide, the length and thickness of the tapered structure portion, the width of the tip portion, the crossing angle of the waveguide, etc. Various modifications can be made without departing from the gist of the present invention.

The schematic diagram explaining the cross type | mold optical waveguide structure by this invention. The perspective view which shows the crossing type | mold optical waveguide structure in 1st Embodiment. FIG. 3 is a cross-sectional view showing a crossed optical waveguide structure in the first embodiment. The figure which shows the mode of the light propagation in the cross type | mold optical waveguide in 1st Embodiment. The figure which shows schematic structure of the optical waveguide circuit in 2nd Embodiment.

Explanation of symbols

1 to 6 Computing core 7 Intermittent optical waveguide 8 Crossing other optical waveguide 9 Tapered structure portion 10 Tapered structure portion 11 Propagating signal light 12 Propagating signal light 13 Lower cladding layer 14 Upper cladding layer 15 Si substrate 16 SiO 2 film Lower clad layer 17 consisting of intermittent optical waveguide 18 optical waveguide 19 taper structure 20 taper structure 21 upper clad layer 22 propagating signal light 23 propagating signal light 24 optical signal transmitting / receiving port 25 optical waveguide 27 optical waveguide crossing portion

Claims (7)

A lower cladding layer formed on the substrate;
A first optical waveguide formed on the lower cladding layer;
One end of the lower cladding layer delimited by the first optical waveguide is formed with its tip directed toward the side surface of the first optical waveguide, and the tip is tapered so that light cannot propagate. A second optical waveguide;
The other end of the lower clad layer delimited by the first optical waveguide is formed with a tip facing the tip of the second optical waveguide, and the tip is tapered so that light cannot propagate. A narrowed third optical waveguide;
A first upper clad layer formed on the lower clad layer and on the tip of the second optical waveguide and having a higher refractive index than the lower clad layer;
An optical waveguide circuit comprising: a second upper clad layer formed on the lower clad layer and on a tip portion of the third optical waveguide and having a refractive index larger than that of the lower clad layer;
The thickness d ₀ of the lower cladding layer is set such that the wavelength of light guided through the first to third optical waveguides is λ, and the refractive index of the lower cladding layer is n ₀ .
d ₀ ≧ λ / n ₀
In relation to
And, wherein the thickness d ₁ of the first and second upper cladding layer, the thickness of the first to third optical waveguide d _2, a refractive index of said first and second upper cladding layer n _{When set to 1} ,
d ₂ ≦ d ₁ ≦ λ / n ₁
An optical waveguide circuit characterized by the following relationship: 2. The optical waveguide circuit according to claim 1, wherein tips of the second and third optical waveguides are at a distance of 30 to 200 μm from a side surface of the first waveguide. The tip portion of the first or second upper clad layer facing the side surface of the first optical waveguide is tapered, and the length of the tapered portion is smaller than 100 μm . 2. The optical waveguide circuit according to any one of 2 above. 4. The optical waveguide circuit according to claim 3 , further comprising a trench region formed by removing a part of the lower cladding layer on both sides of the first or second upper cladding layer. The substrate is characterized by a silicon-on-insulator substrate, the optical waveguide circuit according to any one of claims 1 to 4. 6. The optical waveguide circuit according to claim 5 , wherein the first to third optical waveguides are made of silicon, and the lower cladding layer is made of silicon oxide. Multicore central processing unit and a light waveguide circuit is applied according to any one of claims 1 to 6.